About collect images for calibration.

[xanhug]

About collect images for calibration.

[xanhug]

Current progress: I directly extracted the compressed files downloaded from IBISCape into the map folder, instead of creating a separate folder within the map folder to store them. Now I can see the calibration board.

[xanhug]

Now I have collected some images for calibration and obtained some parameters by running calculate_distortion_parameters.py, but these parameters seem to have nothing to do with 139.1016?

Camera matrix : 

[[214.28235796   0.         799.5       ]
 [  0.         214.14580484 449.5       ]
 [  0.           0.           1.        ]]
Distortion Matrix : 

[[ 2.98099705e-03 -4.35741787e-03 -5.61294835e-05  1.27164886e-05
   1.80059518e-03]]
rvecs : 

(array([[0.22166708],
       [0.23211306],
       [1.58715942]]), array([[0.20972128],
       [0.20671492],
       [1.57471855]]), array([[0.10820797],
       [0.10391412],
       [1.57302881]]), array([[0.50355373],
       [0.48164285],
       [1.50607193]]), array([[0.45873178],
       [0.45762182],
       [1.52581263]]), array([[0.45941141],
       [0.4598727 ],
       [1.52925688]]), array([[0.23872499],
       [0.2616931 ],
       [1.55377912]]), array([[0.45415101],
       [0.44368763],
       [1.53894023]]), array([[0.42452736],
       [0.40566571],
       [1.55872336]]), array([[0.2716295 ],
       [0.28624505],
       [1.58581533]]), array([[0.46011214],
       [0.45866348],
       [1.52386996]]), array([[0.08822307],
       [0.08204556],
       [1.57440487]]), array([[0.00347659],
       [0.03479309],
       [0.00905184]]), array([[0.56930725],
       [0.55490375],
       [1.49517161]]), array([[0.46209434],
       [0.46135466],
       [1.53002564]]), array([[0.19148223],
       [0.18293333],
       [1.57392301]]), array([[0.45955071],
       [0.45793221],
       [1.52471733]]), array([[0.45750716],
       [0.4553597 ],
       [1.52393647]]), array([[0.39585859],
       [0.38174277],
       [1.55457805]]), array([[0.33984159],
       [0.33644849],
       [1.55513569]]), array([[0.10041518],
       [0.09533827],
       [1.57324022]]), array([[0.00378244],
       [0.0888063 ],
       [0.00631415]]), array([[0.23874479],
       [0.2573771 ],
       [1.59453126]]), array([[0.44493193],
       [0.42638195],
       [1.55320535]]), array([[0.36492549],
       [0.358147  ],
       [1.5518794 ]]), array([[0.45911729],
       [0.45845451],
       [1.52741744]]), array([[-0.0466223 ],
       [ 0.05257099],
       [-1.56246822]]), array([[-0.04492518],
       [ 0.33337589],
       [-0.00070372]]), array([[0.00424974],
       [0.00491133],
       [0.01047056]]))
tvecs : 

(array([[ 6.59723144],
       [-4.77955409],
       [ 9.6155787 ]]), array([[ 6.23746009],
       [-4.95289208],
       [ 9.62410831]]), array([[ 4.72545169],
       [-5.01703479],
       [ 9.46942248]]), array([[ 9.7733353 ],
       [-5.18042499],
       [10.18032002]]), array([[ 9.51692199],
       [-4.98669711],
       [ 8.62480518]]), array([[ 9.53951498],
       [-4.9755155 ],
       [ 8.85589948]]), array([[ 5.98607525],
       [-5.0324915 ],
       [13.68420973]]), array([[ 9.42421348],
       [-4.95611648],
       [ 8.21378837]]), array([[ 9.07502163],
       [-4.95155019],
       [ 8.31267258]]), array([[ 7.35116272],
       [-4.70289733],
       [ 9.36909196]]), array([[ 9.50337616],
       [-4.98920092],
       [ 8.37361719]]), array([[ 4.44395759],
       [-5.00953886],
       [ 9.36904114]]), array([[-2.34885378],
       [-5.05964229],
       [ 9.56202647]]), array([[10.86401171],
       [-5.07950019],
       [ 8.08573514]]), array([[ 9.58901652],
       [-4.96930508],
       [ 8.93467486]]), array([[ 5.90826339],
       [-4.99996169],
       [ 9.67476568]]), array([[ 9.50968227],
       [-4.98894112],
       [ 8.49963576]]), array([[ 9.46067274],
       [-4.98651146],
       [ 8.28380494]]), array([[ 8.7098112],
       [-4.9658785],
       [ 8.5301649]]), array([[ 7.99445368],
       [-4.95371473],
       [ 8.92081057]]), array([[ 4.61540717],
       [-5.01409643],
       [ 9.40557943]]), array([[-1.75666834],
       [-5.04612606],
       [ 9.8806514 ]]), array([[ 6.95741384],
       [-4.66148217],
       [ 9.56343023]]), array([[ 9.30779252],
       [-4.94145645],
       [ 8.20729119]]), array([[ 8.31617567],
       [-4.96286955],
       [ 8.76270472]]), array([[ 9.52611465],
       [-4.98191613],
       [ 8.74144602]]), array([[-2.08250646],
       [ 0.94791493],
       [ 9.75200973]]), array([[ 0.32376455],
       [-4.93061337],
       [16.35661429]]), array([[-2.66707133],
       [-5.07063039],
       [ 9.3831467 ]]))

Process finished with exit code 0

So, could you please tell me what is the parameter 139.1016 here?

LIDAR2IMG = {
    "rgb_front":np.array([[800.0, 214.35935394, 0.0, -1200.0], [450.0, 0.0, -214.35935394, -139.10161515000004], [1.0, 0.0, 0.0, -1.5], [0.0, 0.0, 0.0, 1.0]]),

What does it mean? If modify the RGB camera configuration as mentioned above and set undistort to False, do I need to change this parameter? If I don't change it, will it have a significant impact? Because in this case, I really don't know how to modify this parameter.

[jiaxiaosong1002]

@xanhug Hi, LIDAR2IMG is a mapping from ego coordinate system to image cooridnate system: it seems that you has not considered the extrinsics (the position of the camera)?

If you set undistort as False, as in the data loading part, it will use the conversion matrix without considering Distortion matrx. As a result, you do not need to conduct calibration at all (much more simple). All you need is to get the intrinsics and extrinsics matrix (you could use CARLA api in https://carla.readthedocs.io/en/latest/tuto_G_pedestrian_bones/#camera-geometry) and obtain the LiDAR2Img.

I think the calibration would not influence a lot especially with small fov.

[xanhug]

@xanhug Hi, LIDAR2IMG is a mapping from ego coordinate system to image cooridnate system: it seems that you has not considered the extrinsics (the position of the camera)?

If you set undistort as False, as in the data loading part, it will use the conversion matrix without considering Distortion matrx. As a result, you do not need to conduct calibration at all (much more simple). All you need is to get the intrinsics and extrinsics matrix (you could use CARLA api in https://carla.readthedocs.io/en/latest/tuto_G_pedestrian_bones/#camera-geometry) and obtain the LiDAR2Img.

I think the calibration would not influence a lot especially with small fov.

Thanks a lot.